Towards Utilization of Neurofuzzy Systems for Taxonomic Identification Using Psittacines as a Case Study

Table 3

Round 3. Data separated by gender for some species (see Appendix); 3,660 patterns presented for use in training; the learning constraints on the fuzzy sets are as follows: 1: keep their relative order, 2: always overlap, 3: be symmetrical, and 4: intersect at 0.5.

Run	1	2	3	4	5
Num. MF for hue var.	8	16	8	16	16
MF shape	Trapezoidal	Triangular	Trapezoidal	Triangular	Trapezoidal
Aggregation function	Weighted sum	Maximum	Weighted sum	Maximum	Weighted sum
Learning procedure	Best	Best per class	Best	Best per class	Best
Constraints	1, 2	1, 2	1, 2, 3, 4	1, 2	1, 2
Learning rate	0.01	0.1	0.01	0.01	0.01
Max. epochs	500	500	500	1,000	1,000
Min. epochs	10	10	10	20	20
Epochs after optimum	25	25	25	50	50
Correctly classified %	97.02	96.56	96.86	96.58	97.9
Num. rules	1,786	1,785	1,786	1,785	1,935
Mean error	0.324590	0.312022	0.330055	0.339617	0.354098